Novel Topology MC pre-preamp

[analog_guy]

I have been simulating an MC pre-preamp which has a slightly different topology from the usual. It is a no GFB design, where the V- inputs are shorted to gnd. The gain of 24 dB is set by a combination of gm and the drain resistors. Upper and lower halves are shorted together at AC frequencies. This configuration offers two advantages. The first is a substantial (30 dB or better) reduction in distortion. The second is a approx a 3 dB increase in SNR due to the effective paralleling of two correlated signal sources whose noise is uncorrelated. Attached is the active circuit schematic.

Simulated distortion is lower than the noise floor for inputs up to 10 mVPP. The simulated input noise for 3x paralleled 2sj74/2sk170 is 0.28nV/sqrt(Hz)

[ji4m]

Looks interesting, although I don't have enough 2sj74 to build it. Do you simulate with slightly unmatched devices, or assume perfect pairings? I hope this isn't bad form, but would you be willing to share the spice cir file?

[analog_guy]

Originally Posted by ji4m

Do you simulate with slightly unmatched devices, or assume perfect pairings

For the initial simulations I assumed matched devices. The noise floor is not substantially changed due to device mismatch, but the distortion is. Devices of a given polarity need to be matched, but devices of opposite polarity do not. That is why the pos and neg sections are tied together at audio frequencies by the two 0.1 uf caps.

Matching devices for distortion entails matching three parameters: Rs, beta, and Vto. For the devices in question lambda is small enough to not require matching. The generalized matching problem can be addressed by using external passive components to adjust the effective Vto and Rs, leaving only beta, which can be matched by device screening. It may be sufficient to match only Idss. Once I buy a number of devices and do a spice parameter extraction I'll know.

[Joachim Gerhard]

I see 1uF not 0.1uF.

[analog_guy]

Originally Posted by Joachim Gerhard

I see 1uF not 0.1uF.

Good catch. However, the two capacitors form a voltage divider with the gate impedance of the follower jfets. So the corner frequency for either 0.1 or 1.0 uF will still be below the audio range.

[Joachim Gerhard]

Then i whould prefer a 0.1uF because they are awailable in very good quality.

[EC8010]

FETs and MCs don't really mix - take a look at the 1/f noise frequency. On the other hand, distortion really shouldn't be a problem at such low levels.

[analog_sa]

Originally Posted by EC8010

FETs and MCs don't really mix

This statement probably has a solid ground in some textbook but is in direct contradiction with many of the best existing designs. Should owners scrap their Vendettas or just use them with MMs?

[Joachim Gerhard]

In the 2SK170 and 2SK74 1/f corner frequency is quite high at 1 to 4kHz but the rise is very slow.
In praxis the noise can be very low.
I have build several stages with 4 or 8 units in parallel that had excelent noise performance.

[Rod Coleman]

Originally Posted by EC8010

FETs and MCs don't really mix - take a look at the 1/f noise frequency. On the other hand, distortion really shouldn't be a problem at such low levels.

The flicker noise is certainly not a problem for practical designs. The moving coil presents near to a dead short to the gate-source at low frequencies, which doesn't hurt. I use only one 2SK369 (7V, 7mA, shunt cascode) with 11 Ohm in the source for degeneration, and this works perfectly for a Lyra Lydian-Beta. If you get an ear close enough to the tweeter, only some higher frequency noise is perceptible, contrary to the 1/f suspicion.

Just watch out for Igsx (excess gate current) when applying the Toshiba JFETs: it starts at about 8V for practical drain currents, I suspect 12V is too much - even small Igsx current is likely to be noisy.

I would be interested to see any device that is claimed to be better for interfacing a moving coil cartridge than a TOSHIBA large geometry JFET, as shown in this design.

I am assuming that BJTs are out due to the need for coupling capacitors, and although I'd enjoy using a valve, the grid leakage current keeps me away from them - turn-ON pulses of microamp level are hard to avoid (and the Lyra's designer, Jonathan Carr is on record in these pages as declaring these to be a factor in needing to fluxbust). That's before you tackle the noise problem with valves, or their high loop area at the input (electromagnetic 'moment') degrading the external field susceptibility. Enough said about Microphony! MOSFETs are out for certain due to noise level.

What's left? Transformers? I'm not fee-paying enough to buy one that's possibly good enough to try it out. Mr van den Hul has given reason enough why he wouldn't use one, and as a designer of magnetic systems generating this very signal, I'll be content to respect his opinion!

The TOSHIBA 2SK369 and his relatives were designed explicitly for use as "first stage in MC and head amplifier", and I find their low noise, low voltage operation, dc-coupled input, pA input current, high gain AND low output impedance circuit possibilities are all excellent properties when solving the MC input problem. And so long as you can buy thousands of these excellent parts for the cost of a trafo that might begin to compete, I see only one choice!!!